Skip to Content

2018 AI NEXTCon, was jointly held by CSDN (China’s largest developers’ community) and AICamp, from 8th to 11th November, in Beijing, China. Focusing on AI technology and application, the conference was composed of keynotes/workshops/deep dive tech talks, featuring in Image Processing/NLP/Machine Learning Tools & Technologies/Data Analytics/Intelligent Finance and so on. Speakers come from Silicon Valley, AI Unicorns, and Chinese Internet Companies. Luckily, I was attending the conference on behalf of SAP Labs China. Below are the major takeaways, and some hints, from my point of view, for SAP developers.

From strategic perspectives:

Artificial Intelligence should always be business driven

There’s a vivid example by the conference coordinator can elaborate this point.

“Suppose a company wants to bake a cake. A very common error most companies made was, they hired a chef to manufacture an oven, or they hired an electronic engineer to bake a delicious cake. A lot of the courses or textbooks regarding Machine Learning teach you how to manufacture an oven from scratch, but fail to tell baking methods or innovative recipes. While the cases for most companies are they only need the baking methods, i.e., models to solve their business problems.”

Especially for SAP, as a business-driven technology company, AI should always be business driven. We could think more about how to augment our current business scenarios with AI, and some end-to-end scenarios which can leverage AI to solve. AI technologies and their applications, especially in image processing and NLP, have made progressive breakthroughs by academia and internet companies like Google, Facebook, and so on.  However, most of the data we are dealing with is transactional data. There are billions of data been generated and processed on SAP systems, which is a great opportunity for SAP to help our customers utilize their data assets to achieve intelligent enterprise.

Open source is the future of AI development

It is well-known that cloud computing and big data cannot go without open source. Same for AI! There are a lot of open source frameworks/trained models/tools/projects out there for business use. For instance, Tensorflow, Keras, SKllearn, LightGBM, Pandas, Numpy, .etc. It’s great that we, as SAP developers, can take advantage of them, and contribute to the communities by giving feedbacks.

From technical perspectives:

Deep is not necessarily always good

Deep learning has become a hot trend in recent years. Deep learning is able to abstract more information from data, like being able to learn more of the nonlinear correlations. While ‘deep’ is not necessarily always good, machine learning can still be efficient in many aspects.

  • When datasets are not large enough
  • If you expect a more explainable model
  • .etc.

Auto Machine Learning

Traditional machine learning requires onerous data preprocessing and hyperparameters tuning. Deep learning is hard to design. Hyperparameter tuning in deep learning is also very troubled. By using Auto Machine Learning can solve these problems. Let AI design AI models. There are a bunch of open source projects for SAP developers to reference:

  • AutoKeras
  • Adanet (by Google)
  • TransmogrifAI (by SalesForce)

To report this post you need to login first.

1 Comment

You must be Logged on to comment or reply to a post.

  1. Nancy Gu

    Interesting points about machine learning, AI and SAP.

    Hope SAP can help customer to get more success by machine learning and AI.

    Great opportunity and challenge for SAPer~

     

    (1) 

Leave a Reply